gf-agent-contributor

Skill: gf-agent-contributor

Overview

This skill provides instructions on how to correctly contribute knowledge, best practices, or API parity warnings to the gf_agent so that it learns how to write better Google Apps Script automation code.

The Architecture

To prevent Git merge conflicts between multiple developers, the monolithic gf_agent/SKILL.md file is auto-generated. You MUST NOT edit gf_agent/SKILL.md directly.

Instead, gf_agent uses a modular knowledge base. The builder.js script aggregates all markdown files in the gf_agent/knowledge/ directory and appends them to a base template (gf_agent/scripts/SKILL.template.md).

How to Contribute Knowledge

When you need to teach gf_agent something new, you must first ensure you are working from a local Git clone, NOT the cached Gemini skills folder.

1. Clone the gf_agent Environment (Sparse Checkout): Because gf_agent is a subfolder of the larger gas-fakes repository, you should use Git's Sparse Checkout feature to download only the gf_agent directory.

   🤖 Agent Instruction: If the user is not yet in a configured workspace, offer to run the sparse checkout setup for them! If they agree, use run_shell_command to execute this exact sequence (adjust the parent directory if they specify one):

   mkdir gf_agent_standalone && cd gf_agent_standalone
   git init
   git remote add origin https://github.com/brucemcpherson/gas-fakes.git
   git config core.sparseCheckout true
   echo "gf_agent/*" >> .git/info/sparse-checkout
   git pull origin main
   gemini skills link ./gf_agent
   

   (If the user prefers to do it manually, provide them the script above).

2. Create a Modular File: Navigate to the gf_agent/knowledge/ directory inside your new standalone environment. Create a new markdown file for your topic.

   • Use a numeric prefix to ensure it sorts correctly during compilation if order matters (e.g., 06-new-feature.md). If it is a standalone topic, any descriptive name is fine.